ICES CM 2000/K:30
Theme Session K
Testing the effects of vessel, gear and daylight on catch
data from the International Bottom Trawl Survey in the
North Sea
Jacques Rivoirard
A number of vessels participate regularly to the IBTS in the North Sea. Because they survey
different areas with different fish densities, the vessel effect has been investigated by selecting
neighbouring trawl stations coming from each given pair of vessels. A Student test is then
performed to see if the observed difference between those two vessels is significant given the
variability of catches.
This analysis has been conducted for cod, haddock and herring at ages 1, 2 and 3+ for each of
quarters 1 and 3 from 1991 to 1997. The most striking result is the repeated significantly
different level of catches by Scotia 2 used in quarter 3 (lower for what concerns cod and
haddock, but not necessarily for herring). Those systematic differences are likely to be due to
the use of a gear which is different from the other vessels.
The effect of daylight on cod, haddock, whiting and herring at ages 1, 2 and 3+ in quarter 1
has been investigated by selecting neighbouring pairs of stations from the same vessel. A
Student test has been performed to evaluate if catches at night are significantly different from,
and for instance lower than, catches at day. The significance of the correlation of catches with
sun elevation has also been assessed. By repeating the analysis for each year from 1983 to
1997, a positive effect of daylight on haddock and herring all ages and on cod 2 and 3+ has
been evidenced. However, obtaining a response function on sun elevation from catch data, is,
despite their number, problematic.
Jacques Rivoirard, Centre de Géostatistique, Ecole des Mines de Paris, 35 rue Saint-Honoré,
F-77305 Fontainebleau-Cedex, France, E-mail: rivoi@cg.ensmp.fr
1
1 Introduction
The International Bottom Trawl Survey (IBTS) in the North Sea is a coordinated multivessel survey. It has been conducted in the North Sea in the 1st quarter of the year since the
mid 1960’s and in the 3rd and other quarters more recently. The main objective of the IBTS is
to provide recruitment estimates and tuning data for ICES (International Council for the
Exploration of the Sea) assessments of several commercially important fish species. However,
standard abundance indices by age group are routinely calculated in a way that does not
account for differences in catchability between vessels, or in catch rates between day and
night.
A number of analyses have been made on these IBTS data (see for instance those referred
to in Rivoirard and Wieland, 2000). The purpose of the present analysis is to test the existence
of a vessel, gear, and daylight effects in these data. This is done by comparing directly the
level of catches for different species at age. However the different vessels cover different
areas. Similarly, the effect of daylight, if any, is expected to be larger in the data from the
North of the North Sea in winter, where the short duration of day results in a high proportion
of hauls performed at night. Thus, a selection of data has to be made prior to test the
differences of catches between vessels or according to daylight.
2 Material
The data used in the present study have been extracted from the ICES IBTS Database for
quarter 1 in 1983-1997 and quarter 3 in 1991-1996. They include in particular agedisaggregated catches (in numbers per hour trawling), vessel, gear, shooting position, date,
time of day and a day/night code. They have been complemented by the local time of day
(used in the present analysis) and the sun elevation. The location of hauls is illustrated in 1991
in the example of quarter 1 (Fig. 2), and in 1996 in the example of quarter 3 (Fig. 1). There
are typically 400 hauls in a quarter 1 survey, and 300 in a quarter 3 one. Species under
consideration are: haddock, cod, whiting, and herring. Three age classes (1, 2, and 3+) are
used for each species (for herring, the class 1, 2, or 3+ represents the number of rings, equal to
age +1).
2
3 Effect of vessel and gear
A number of vessels participate to each survey (Table 1). The vessel effect has been
investigated:
- for quarter 1 in years 1991 to 1997 on haddock and cod 2 and 3+ and on herring 1, 2 and 3+;
- for quarter 3 in years 1991 to 1996 on haddock, cod, whiting and herring 1, 2 and 3+.
Fig. 1 shows the spatial distribution of the hauls made by each vessel in the example of
quarter 3 in 1996.
3.1 Selection of data from overlapping pairs of vessels
Because the different vessels do not cover the same area, the differences between their mean
catch level are not significant of a vessel effect. However there are some overlappings
between the areas covered by the different vessels. This has been exploited in comparing
directly the catches between any pair of vessels which are “overlapping” in space. Rather than
delineating the intersection between the areas covered by the two vessels, we have selected
(for each quarter and each year), the closest pairs of hauls (referred to as duplicates) from the
two vessels. The limit distance has been taken as 20 n.m. (increasing this would increase the
number of selected hauls, but also the risk to extend this 2x2 vessels comparison to vessels
that do not really overlap). Care was taken that one haul is not selected more than one time for
a given pair of vessels. The number of spatially overlapping pairs of vessels is the following:
1991
1992
1993
1994
1995
1996
1997
total
Quarter
1
3
16
4
15
10
15
9
11
6
10
3
10
8
9
86
40
Overlapping is more frequent in quarter 1 than quarter 3, due to the higher number of vessels
involved.
3
3.2 Statistical test
For each pair of overlapping vessels (and each survey), a number of duplicate hauls is then
selected. However the mean difference between the catches (equal also to the mean of the
differences between duplicates) can be high without being significant (e.g. due to a few
number of duplicates, or to a locally high variability). It can be assumed that the continuous
component of the spatial structure is the same within a duplicate, and so is filtered out from
the difference of a duplicate. Moreover all the samples are different. So the differences
between each duplicate can be assumed as independent. A Student t-test is then performed on
these duplicate differences to see if their mean is significantly different from 0, that is, if the
means for the two vessels are significantly different (while the sample values are theoretically
assumed to be normally distributed, the test is known to be robust against departure from
normality).
An observed mean difference is considered as significantly different from 0 at probability
95% if, under the starting hypothesis that the mean is theoretically 0, it corresponds to a
probability p smaller than the risk 5% = 1/20. This risk represents the probability to consider
the difference as significant and to reject the starting hypothesis, when in fact this is true.
When the test is repeated (different pairs of vessels, different years), it is normal, if the mean
is theoretically 0, to observe some results considered as significant and the risk gives their
expected proportion. So, if there is no vessel effect, we should expect, on average over all
years and for each species at age, 86/20, that is about 4 significant differences in quarter 1,
and 40/20 = 2 significant differences in quarter 3.
3.3 Results
The pairs of overlapping vessels presenting differences significant at probability 95% are
listed in Table 2.
In quarter 1, the number of significant differences is small (in particular there is none for
haddock 3+), and generally smaller than its expected value of 4. In addition no repeatable
feature is observed, except maybe the appearance of WAH2 or 3.
In quarter 3 on the contrary, the number of significant differences is high, particularly for
younger cod and haddock (ages 1 and 2) and older herring (2 and 3+), though there are no
4
significant difference for cod 3+, and only 1 for herring 1. The most striking result concerns
the nearly systematic presence of vessel SCO2 in the incriminated pairs. (This is not due to a
higher representation of SCO2: considering the 40 overlapping pairs of vessels over all years
for quarter 3, CIR contributes: 20 times; SCO2 and TRI2: 17 each; THA: 11, THA2, ISI and
WAH2: 4 each; and WAH3: 3.)
When it is significantly different, the level of capture of SCO2 is always lower than other
vessels for what concerns cod and haddock:
SCO2 < CIR in 1991 (cod 1-2), 1992 (cod 1, haddock 1-2-3+), 1993-94 (cod and haddock 1
and 2), 1995 (cod and haddock 1), 1996 (cod 2)
SCO2 < WAH2 in 1992 (cod 1, haddock 1, 2 3+)
SCO2 < TRI2 in 1991 (haddock 1-2), 1992 (cod 2, haddock 2-3+), 1995 (haddock 1-2)
SCO2 < THA2 in 1996 (cod 1)
However, for herring, while all significant differences involve SCO2, its level can be higher
or lower:
SCO2 < CIR in 1992 and 1995 but > CIR in 1991 (herring 2 and 3+)
SCO2 < WAH2 in 1992 (herring 3+)
SCO2 < TRI2 in 1993 (herring 2) but > TRI2 in 1995 (herring 1 and 2)
3.4 Conclusion on vessel and gear effect
Significant differences are repeatedly observed for Scotia 2 in quarter 3. Interestingly, the gear
used in the data analysed is a GOV (chalut à Grande Ouverture Verticale) in all cases, except
in quarter 3 for CIR in 1991 (using a Granton Trawl) and for SCO2 (using a Dutch Herring
Trawl 48 feet in 1991 and an Aberdeen 48 feet Trawl from 1992 to 1996). So a vessel effect
appears when there is a difference in gear, while no vessel effect appears when using the same
gear. The vessel effect appears to correspond to a gear effect.
3.5 Ratio of catches between vessels
Assuming that, due to the gear employed, a vessel effect exists for Scotia 2, ratios of catches
for Scotia 2 versus all other vessels in quarter 3 have been estimated. Nearest pairs of hauls
without replication between Scotia 2 and each other vessel, with limit distance 20 n.mi., have
been selected. A Scotia 2 haul can have for instance 2 neighbours coming from 2 different
vessels, then making two pairs and so being counted twice. For a given year, the ratio is the
5
ratio between the mean of Scotia 2 catches and the mean of other vessels catches, from
selected pairs. For all years together, the ratio is the mean of the ratios for each year - not the
ratio between the means from all pairs.
Ratios of catches between SCOTIA2 and other vessels in quarter 3:
year
pairs
Cod1
Cod2
Cod3p Had1
Had2
1991
1992
1993
1994
1995
1996
all
58
88
60
50
48
42
0.11
0.13
0.38
0.28
0.20
0.57
0.28
0.38
0.37
0.60
0.37
0.53
0.25
0.42
0.74
1.31
0.86
0.61
0.68
0.62
0.80
0.46
0.25
0.56
0.35
0.56
0.90
0.51
0.39
0.31
0.63
0.37
0.45
0.95
0.52
Had3p Herr1 Herr2 Herr3
p
0.56
1.36
0.54
0.40
0.25
0.20
0.05
0.06
0.55
0.08
0.05
0.04
0.48
0.25
0.12
0.06
0.72
0.20
0.10
0.09
0.71
0.17
0.35
0.49
0.54
0.38
0.20
0.19
4 Effect of daylight
In the present analysis, daylight for each haul is accessible indirectly through the day-night
code, the local time of day, and the sun elevation. Hauls are preferentially performed at day,
but in winter days are short and a number of hauls are at night (essentially in the North of the
North Sea), thus spanning a large range of daylight. The portion of N-hauls, as well as their
location, vary with year, and so a postulated daylight effect is likely to bias unevenly the
survey based abundance indices.
In the present analysis, the effect of daylight has been investigated on the IBTS data from
quarter 1, years 1983-1997, for haddock, cod, whiting and herring 1, 2 and 3+.
4.1 Selection of data
Because N-hauls are not located evenly within the North Sea, they are not directly comparable
to D-hauls. To span D/N conditions under otherwise similar conditions, hauls have been
selected to include if possible each N-haul as well as the closest D-haul from the same year,
same vessel, and at distance less than a given limit. The number of selected hauls varies with
this limit distance as follows:
10
20
30
40
50
60
70
72
610
1342 1540 1560 1560 1560
6
nm.
As each N-haul and its close D-haul both belong to the area covered by the same vessel, a
rather large limit of 50 nm has been selected, ensuring also to include a high number of data
(780 out of a total of 1003 N-hauls in quarter 1 are then selected, those far from a D-haul are
not). Such selected hauls are visualized in Fig. 2 in the example of 1991. The majority are in
the Northern part of North Sea, however there are also a number of these along the German
coast in 1983-1987 and 1989. Only a few are in Skagerrak and Kattegat, where cod can be
abundant.
On the contrary to the analysis of the vessel effect (where the selected data were paired, in
fact only the difference of each duplicate was used in the test), the data selected here have not
been paired for tests and are used as a single set to analyse the effect of D/N as well as this of
time of day and sun elevation. Because the existence of a spatial structure (see for instance
Rivoirard and Wieland, 2000), such data are not independent but their degree of dependence
is low considering the amplitude of nugget effect or short component in the variogram.
4.2 Day/Night effect
The ratio between the mean catch of night hauls and this of day hauls, plotted for each species
at age through years, appears to be frequently less than 1 (Fig. 3). However such results are
based on a limited number of data, often with a high variability. A Student test has been used
to determine where the N-mean is significantly different from (and then for instance lower
than) D-mean (the test assumes that the data are independent and that D-data and N-data are
normal with the same variance, but it is known to be robust against departure from normality).
We start from the hypothesis Ho: n = d (N-mean = D-mean, i.e. no effect), and reject it when
the difference between the observed means is too large, i.e. corresponds to a probability p too
small compared to a given risk (taken here as 2/15 = 0.133). So if p < 2/15, Ho is to be
rejected (the risk is the probability to reject Ho although Ho is valid). Then according to
whether n < d or n > d, the hypothesis H1: n < d or H-1: n > d is to be preferred.
Detailed results are listed in Table 3, in the example of haddock age 2 (see also Fig. 3 for a
proportional representation of the 1-p values).
7
If Ho is true, it should be rejected on average (but no necessarily exactly) for 2 years out of
15, and each of H1 and H-1 preferred for 1 year. This has lead us to adopt, for a given species
at age, the hypothesis H0, H1, or H-1 appearing in an excessive number of years compared to
expected, and considering also all years together.
Such summarized results are listed in Table 4. A D/N effect (H1: n < d) is evidenced for
haddock and herring all ages, cod 2 and 3+. In other cases, Ho (n = d, no effect) is not
rejected.
4.3 Effect of time of day and sun elevation
The response of catches on time of day and sun elevation has firstly been searched by plotting
the experimental regressions. However no clear quantified relation appears, even in the best
cases like haddock age 2 (Fig. 4 and 5) and even when considering all years together (Fig. 6).
The relation of sun elevation on time of day can be reasonably well represented by a cosine
function centered at noon (Fig. 7). As a consequence, a cosine response of catches on time
would closely correspond to a linear response on sun elevation. While sun elevation makes a
better distinction between day and night than time of day does, its value at noon decreases
with latitude and increases during winter. A question like whether the supposedly maximal
level of catches at noon varies or not with sun elevation, e.g. with the julian day, while it may
be important in the view of correcting catches from a daylight effect, seems out of range from
the data.
While the response function of catches to sun elevation is difficult to quantify, we have tested
the dependence by determining whether the correlation is significantly different from 0 (and
for instance is > 0). The correlation r measures the strength of the linear dependency, more
exactly the strength of the linear part of dependency. In our case, the dependency is not
necessarily linear, but it is expected to be monotonic with a positive correlation, if an effect
exists.
We start from the hypothesis Ho: r = 0 (no effect). Then, when the number n of independent
samples is large enough (> 30; in our case there are about 100 selected samples each year,
although not necessarily independent), the estimated correlation r* is normally distributed
8
with mean 0 and variance 1/(n-1). Ho is to be rejected when r* deviates too much from 0, i.e.
corresponds to a too small probability. We will consider a risk of 2/15, that is, 1/15 for each
sign of correlation. Hypothesis H1: r > 0 is to be preferred when r* > 1.5011/√(n-1), since the
probability of this event under Ho is P(r* > 1.5011/√ (n-1)) = 1/15 = 0.067. Similarly H-1: r <
0 is to be preferred if r* < -1.5011/√ (n-1). On average, if Ho is true, it should then be rejected
for 2 years out of 15, and each of H1 and H-1 preferred for 1 year.
Table 5 gives the detailed results in the example of haddock age 2, while summarized results
are listed in Table 6.
These results match those on N- and D-means. The correlation is significantly positive where
the N-mean is significantly lower than D-mean, that is, for haddock and herring all ages, cod 2
and 3+. No significant effect is evidenced in other cases (except maybe H-1, a negative
correlation of cod1 with sun elevation).
4.4 Conclusion of tests on daylight effect
An effect of daylight (catch level lower at night, positive correlation between catch and sun
elevation) is evidenced for haddock and herring all ages, cod 2 and 3+. No effect is visible for
cod 1 and for whiting all ages.
Acknowledgements
The European Union provided financial support (DG XIV study No 97/0009, MIQES – The
use of multivariate data for improving the quality of survey-based stock estimation in the
North Sea).
Reference
Rivoirard J. and Wieland K., 2000.Correcting daylight effect in the estimation of fish
abundance using kriging with external drift, with an application to juvenile haddock in
North Sea. ICES CM 2000/K: 31
9
Table 1: List of vessels participating to IBTS, with their code in the present analysis and their
year of use in quarters 1 (1991-1997) and 3 (1991-1996).
ARG = Argos
DAN2 = Dana (new)
ISI = Isis
JHJ = Johan Hjort (new)
SCO2 = Scotia (new)
THA = Thalassa
TRI2 = Tridens (new)
WAH2 = Walther Herwig
GOS = G.O. Sars
SOL = Solea
MIC = Michael Sars
WAH3 = Walther Herwig III
THA2 = Thalassa (new)
CIR = Cirolana
JOH = Johan Hjort (old)
quarter 1
ARG DAN2 83-97
SCO2 85-97
TRI2 91-97
THA 83-96 THA2 97
WAH2 87-93 WAH3 94-97
MIC 93 95-97
ISI 84-87 89-91 93
GOS 92&94
JHJ 91
SOL 92
quarter 3
ARG CIR SCO2 TRI2 91-96
ISI 91&93
THA 92-94
THA2 96
WAH2 92
WAH3 96
10
Table 2: Pairs of overlapping vessels with difference in mean catch significant at probability
95% (see details in text). The number of pairs of hauls considered and the ratio between the
mean catch of the 1st vessel on this on the 2nd vessel are indicated.
Quarter Species at age year
Nb pairs
Ratio
1st vessel
2nd vessel
1
1
1
cod 2
cod 2
cod 2
1991
1996
1997
34
15
33
0.57
0.25
0.5
SCO2
TRI2
WAH3
WAH2
DAN2
MIC
1
1
1
1
cod 3+
cod 3+
cod 3+
cod 3+
1992
1995
1995
1997
3
26
17
33
0.21
0.54
0.43
0.41
WAH2
MIC
WAH3
WAH3
TRI2
WAH3
SCO2
MIC
1
haddock 2
1996
23
0.53
SCO2
WAH3
1
1
herring 1
herring 1
1993
1996
15
12
0.29
0.17
THA
THA
DAN2
DAN2
1
1
1
1
herring 2
herring 2
herring 2
herring 2
1991
1994
1995
1997
14
35
2
21
0.26
0.099
0.037
0.39
DAN2
SCO2
THA
TRI2
WAH2
WAH3
WAH3
THA2
1
herring 3+
1992
11
0.18
GOS
SCO2
3
3
3
3
3
3
3
3
cod 1
cod 1
cod 1
cod 1
cod 1
cod 1
cod 1
cod 1
1991
1992
1992
1992
1993
1994
1995
1996
38
36
34
12
34
32
33
4
0.08
0.15
0.12
0.6
0.23
0.34
0.28
0.17
SCO2
SCO2
SCO2
THA
SCO2
SCO2
SCO2
SCO2
CIR
CIR
WAH2
TRI2
CIR
CIR
CIR
THA2
3
3
3
3
3
cod 2
cod 2
cod 2
cod 2
cod 2
1991
1992
1993
1994
1996
38
14
34
32
32
0.52
0.31
0.49
0.59
0.27
SCO2
SCO2
SCO2
SCO2
SCO2
CIR
TRI2
CIR
CIR
CIR
3
3
3
3
3
haddock
haddock
haddock
haddock
haddock
1991
1992
1992
1992
1993
21
36
33
34
34
0.26
0.43
0.45
0.21
0.62
SCO2
SCO2
CIR
SCO2
SCO2
TRI2
CIR
WAH2
WAH2
CIR
1
1
1
1
1
11
3
3
3
3
haddock
haddock
haddock
haddock
1
1
1
1
1994
1995
1995
1995
32
33
8
12
0.49
0.58
0.36
0.31
SCO2
SCO2
CIR
SCO2
CIR
CIR
TRI2
TRI2
3
3
3
3
3
3
3
haddock
haddock
haddock
haddock
haddock
haddock
haddock
2
2
2
2
2
2
2
1991
1992
1992
1992
1993
1994
1995
21
36
14
34
34
32
12
0.35
0.23
0.32
0.23
0.6
0.61
0.63
SCO2
SCO2
SCO2
SCO2
SCO2
SCO2
SCO2
TRI2
CIR
TRI2
WAH2
CIR
CIR
TRI2
3
3
3
haddock 3+
haddock 3+
haddock 3+
1992
1992
1992
36
14
34
0.18
0.32
0.3
SCO2
SCO2
SCO2
CIR
TRI2
WAH2
3
herring 1
1995
12
0.051
TRI2
SCO2
3
3
3
3
3
herring 2
herring 2
herring 2
herring 2
herring 2
1991
1992
1993
1995
1995
38
36
21
33
12
0.62
0.07
0.046
0.068
0.021
CIR
SCO2
SCO2
SCO2
TRI2
SCO2
CIR
TRI2
CIR
SCO2
3
3
3
3
herring 3+
herring 3+
herring 3+
herring 3+
1991
1992
1992
1995
38
36
34
33
0.55
0.053
0.043
0.057
CIR
SCO2
SCO2
SCO2
SCO2
CIR
WAH2
CIR
12
Table 3: Day/night effect in the example of haddock age 2. Student test at risk 2/15 for Daymean and Night-mean:
n/d = ratio between Night-mean and Day-mean;
p = corresponding probability if no effect;
H = resulting hypothesis;
H = 0 (n = d) if p > 2/15 = 0.133;
H = 1 (n < d) if p < 2/15 and n < d;
H = -1 (n > d) if p < 2/15 and n > d.
n/d
p
H
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
All years together
0.64
0.46
0.39
0.81
0.87
0.62
0.8
0.21
0.29
1.22
1.13
0.63
0.6
0.82
0.6
0.67
0.14
0.11
0.01
0.59
0.7
0.21
0.5
0.02
0.01
0.64
0.72
0.3
0.2
0.49
0.32
0
0
1
1
0
0
0
0
1
1
0
0
0
0
0
0
1
Table 4: Day/night effect. Testing H1 (n<d, i.e. Night mean < Day mean) and H-1 (n>d)
under Ho (n=d). Number of years for each hypothesis ("all" means "all years together").
Species
age
Ho
H1
H-1
conclusion
Expected
13
1
1
Haddock
1
2
3+
7
11
8
8+all
4+all
7+all
0
0
0
H1
H1
H1
Cod
1
2
3+
14+all
11+all
11
1
4
3+all
0
0
1
H0
H1
H1
Whiting
1
2
3+
15+all
11+all
15+all
0
3
0
0
1
0
H0
H0
H0
Herring
1
2
3+
14
13
12
1+all
2+all
3+all
0
0
0
H1
H1
H1
13
Table 5: Test of correlation of catches with sun elevation in the example of haddock age 2:
Corl = correlation;
Limit = absolute limit value for significance at risk 2/15;
H = resulting hypothesis:
H = 0 for correlation not significantly different from 0;
H = 1 for correlation significantly > 0;
H = -1 for correlation significantly < 0.
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
all
corl
limit
H
0.12
0.17
0.21
0.08
0.09
0.09
0.07
0.31
0.29
-0.07
-0.16
0.1
0.09
0.07
0.04
0.08
0.14
0.15
0.12
0.15
0.14
0.16
0.14
0.18
0.16
0.15
0.15
0.17
0.16
0.15
0.14
0.04
0
1
1
0
0
0
0
1
1
0
-1
0
0
0
0
1
Table 6: Correlation of catches with sun elevation. Testing H1 (r>0) and H-1 (r<0) under
Ho (r=0) for correlation r between catch and sun elevation.
Species
age
Expected
Ho
H1
H-1
13
1
1
conclusion
Haddock
1
2
3+
8
10
9
7+all
4+all
6+all
0
1
0
H1
H1
H1
Cod
1
2
3+
10
10+all
12
2
4
3+all
3+all
1
0
H0
H1
H1
Whiting
1
2
3+
14+all
11+all
13+all
0
2
1
1
2
1
H0
H0
H0
Herring
1
2
3+
11
13+all
11
4+all
2
4+all
0
0
0
H1
H1
H1
14
54
56
58
60
62
1996 quarter 3
50
52
ARG
CIR
SCO2
THA2
TRI2
WAH3
-5
0
5
10
Fig. 1: Location of trawl stations and vessels in the example of quarter 3, 1996
15
50
52
54
56
58
60
62
IBTS 1991 quarter 1
-5
0
5
10
pairs of close day/night samples from same vessel
Fig. 2: Locations of hauls in the example of quarter 1, 1991: day-hauls (open circles); nighthauls (black circles); lines between close pairs of night and day hauls from same vessel.
16
haddock2
haddock3p
1.0
0.5
0.0
0.0
0.0
0.4
0.5
0.8
1.0
1.5
1.2
1.5
haddock1
84 86 88 90 92 94 96
84 86 88 90 92 94 96
cod1
cod2
cod3p
0.0 0.5
0
0
2
1
4
2
6
3
8
1.0 1.5 2.0 2.5
84 86 88 90 92 94 96
84 86 88 90 92 94 96
84 86 88 90 92 94 96
whiting1
whiting2
whiting3p
2.0
0.0
0
0.0
1
0.5
1.0
2
1.0
3
1.5
4
2.0
3.0
84 86 88 90 92 94 96
84 86 88 90 92 94 96
84 86 88 90 92 94 96
herring1
herring2
herring3p
84 86 88 90 92 94 96
0
0
0.0
2
2
0.4
4
4
6
0.8
6
8
8
1.2
10
84 86 88 90 92 94 96
84 86 88 90 92 94 96
84 86 88 90 92 94 96
Fig. 3: Ratios between mean of night catches and mean of day catches plotted through years
for the different species at age in quarter 1. Symbols are proportional to 1-p (see details in
text): the larger the symbol, the more significantly different from 0 the difference in means.
17
600
1983
1984
•
•
•
•
15
20
•
0
•
•
5
10
•
15
20
10
•
•
15
•
0
0
20
0
•
•
• • •
10
15
20
1400
•
•
•
15
20
•
0
5
•
•
•
10
•
15
20
•
•
•
•
•
•
15
20
0
5
•
10
15
•
•
0
•
•
•
•
•
5
•
10
20
1997
•
•
•
•
•
0
•
•
400 800
300
0 100
• •
20
•
•
•
•
5
•
•
1400
•
0
•
•
15
•
1996
•
•
•
10
1995
• •
10
400
10
• ••
0
•
•
5
5
200
0
0
•
•
1994
1500
•
• •
•
0
•
•
20
•
•
0 500
400 800
•
15
•
1993
•
•
10
• • •
1991
•
1992
•
5
600
••
5
20
1500
0
• •
400
200
40
•
10
20
•
•
5
15
200
•
• •
0
•
•
0
5
400 600
•
•
•
•
•
0
15
•
1990
•
•
10
•
•
1989
•
5
•
•
•
0
10
0
1988
•
0 50
5
• • • •
•
•
500
•
•
0
20
•
•
400
150
• •
•
15
•
1987
200
250
1986
•
•
•
•
•
0
10
•
0
• • •
0
5
500 1000
600
•
•
•
500 1000
•
•
•
15
20
•
•
•
•
•
•
0
200
•
0 200
400
•
•
0
80 120
1985
0
5
10
15
20
Fig. 4: Experimental regression of catches on time of day through years in quarter 1, in the
example of haddock 2.
18
1983
1984
1985
•
•
• •
0
20
-40
•
-20
1986
0
•
600
0
•
•
20
-40
• •
•
-20
0
0
20
•
•
-40
• ••
-20
•
0
•
-40
-20
20
•
• • • • • •• •
••
• •
•
-40
-20
600
400
0
20
-20
0
6000
20
0
20
•
•
•
•
•
••• • •• •
•
•
-40
-20
0
•
•
• •
-40
•
20
•
2000
•
0
200
•
• •
• •• • •
•
1997
•
0
•
•
1994
•
-20
•
1996
•
-40
• •
•
-40
2000 4000
•
•
0
••
•
•
1995
•
•
20
0
2000
•
0
• • •
•
•
•
6000
1500
•
0
500
•
20
•
400
•
•
•
0
•
•
•• •
-20
1993
•
••
-40
•
•
•
• •
•
200
•
•
•
•
0
0 100
••
1992
•
20
1991
•
•
-20
0
•
20
300
-40
•
••
•
•
•
•
150
0 50
•
• •
400 800
•
•
•
•
-20
1990
•
• •
•
0
•
•
•
•
1989
•
-40
•
• •
200
0 100
• • •
•
•
1988
•
•
•
250
20
•
• •
•
•
0
300
•
• •
-20
•
• •
•
•
•
-40
•
1987
•
•
• • • •
• •
•• • •• • •
•
•
-20
0
20
400 800
-20
•
•
0
-40
•
•
0
••
•
•
•
0
•
•
•
0
••• •
•
400
200
•
•
•
400 800
800
600
•
•
••
-40
•
• •• •
-20
•
•
•
0
•
20
Fig. 5: Experimental regression of catches on sun elevation through years in quarter 1, in the
example of haddock 2.
19
0
2000 4000 6000 8000
maeg2
0
5
10
15
20
800
time
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
0
•
•
•
0
•
•
•
•
R2 0.004 •
200
400
600
•
5
10
15
20
0
2000 4000 6000 8000
time
-40
-20
0
20
sun elevation
•
•
•
•
•
•
•
R2 0.004
•
•
•
•
•
•
•
•
•
0
200
400
600
•
-40
-20
0
•
20
sun elevation
Fig. 6: Catches vs. time of day and sun elevation for all years together in quarter 1, in the
example of haddock 2: plots of values and experimental regressions.
20
20
-40
sun elevation
-20
0
R2
.. ..
.
.
0
. .. .
.
. ... ......................... .
.. ... ..... .
.. .......................................................................................
.............................................................................................................................................
. ....................................... ..
.................................................................................................................................................................................. .
.
.................................................................................................................................................................
.
.
.
.
.
.
.
...................................... ..
.
.
.
.
..
0.923
. ........................................................ ..
. .........................................
............................. .
.............................................
. ..............................
...................................................
...........................
.
.
.
.
.
.
.
.......................
.............................
. ..... .
...........................
.................
............. .
...................
.
..............
...... ....
..... .
.................. . .
. ..........
.
.......
.............. .
.............. .
.. . ..
...........
.
..
........... .
..
................ . .
.
.
... .
.........................
.
.
.. ......... . .
.... .
... .. ..
.
. . ...
..
.
.
5
10
15
20
time
Fig. 7: Sun elevation on time of day for quarter 1 data (fitted by least squares with a cosine
function maximal at noon: -11.63+31.16 cos(2π (t-12)/24))
21